Space Qualification for Semiconductor Devices

Transcription

1 Space Qualification for Semiconductor Devices Sammy Kayali Jet Propulsion Laboratory California Institute of Technology (818) May 14, 2007 Acknowledgment: The work described in the paper was conducted at the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. 1

2 Tutorial Objectives Describe the environmental, reliability and operational challenges for space applications. Provide a description of space qualification requirements and associated product testing and evaluation. In addition, semiconductor device processing, test and characterization requirements along with supporting data and application support will also be presented. Lastly, practical considerations for development of qualification plans for space applications will also be presented. 2

3 Outline Challenges for Space Applications Space Qualification Requirements Reliability Considerations Radiation Effects Technologies and Failure Mechanisms Planning and Conducting Qualification Tests 3

4 Space Application Challenges

5 Space application Challenges The application of microelectronics in space systems faces a number of challenges Space environment Reliability and Lifetime Lack of accessibility Consideration must be given to the impact of: Lot non-uniformity & traceability Temperature range of component Availability of test data It is critical that all aspects of reliability and relevant known failure modes and mechanisms be addressed prior to the insertion of the component in the application 5

6 Environmental Challenges for Space Applications Launch Environment Operation at Temperature Extremes Thermal Dissipation Space Radiation Effects Spacecraft Charging Micrometeoroids 6

7 THE ULTIMATE SOURCE OF SPACECRAFT INTERACTIONS: 7

8 Reliability Challenge Determine if a component, designed for a limited life terrestrial environment, is suitable for a long life application in a space environment. Constraints: Limited information on the product test data, Design, Materials or limitation in performance. Must Arrive at an acceptable solution under cost and schedule constraints Available test sample size is limited Utilize all available information on test structures or process controls 8

9 Challenges We Face Now Industry challenges Diminishing Hi-Rel suppliers The space and military infrastructure no longer drives the semiconductor industry Little industry support for small volume procurements & space applications Environmental Challenges Operation at extreme environments Reduction in available mass and volume Radiation effects Technical Challenges Smaller feature sizes and miniaturization Low power devices and designs Thermal dissipation New materials and processes Increased device complexity 9

10 Industry Trends Smaller Feature Size Requires higher resolution analysis and sample prep equipment Increased sensitivity to radiation effects More complex devices Require improved and faster tester capability Fewer providers of Hi-Rel components Necessitates increased use of COTS Shift in emphasis from device specification to device analysis Accelerating technology development trend Increased performance gap between State-of-the-Art and Hi-Rel suppliers Faster introduction of new device technologies 1.0µm Yesterday < 0.10µm Today 10

11 Parts Temperature Application Range 4 Parts Application COTS Military Temp. Range S/C Bus Mars Surface Apps. Detector Apps Temperature (C) 11

12 The Relative Cost of IC Failure Relative Cost of a Integrated Circuit Failure (Rule of Thumb) 1X 10X > 100X At Chip Manufacturer Low Impact to System Reliability At End-product Manufacturer Medium to High Impact to System Reliability In the Field Catastrophic Impact To System Reliability 12

13 Space Qualification Requirements

14 Space Qualification Qualification can be defined as the verification that a particular component s design, fabrication, workmanship, and application are suitable and adequate to assure the operation and survivability under the required environmental and performance conditions. Industry and Military Standards are designed with the objective of verifying process controls and product quality A methodology for qualification based on continual interaction between the device manufacturer and the user is the desired state. Process Qualification Product Qualification Product Acceptance Mars Exploration Rover,

15 Process Qualification Procedures and data demonstrating the control of the entire process of design and fabrication for a particular technology. Includes all aspects of the process: Acceptance of starting materials Documentation of procedures Implementation of handling procedures Establishment of Lifetime and failure data Test Structures Process Monitors Technology Characterization Vehicles Etc. Utilize the manufacturer existing and defined process with established reliability and qualification procedures and practices 15

16 Product Qualification Verification that a component will satisfy the design and application requirements under specific conditions. Design Verification Verification of model or simulation and layout prior to fabrication Requires detailed knowledge of the design tools, device physics, layout tools, fabrication and test. Product Characterization Determine an understanding of the limitations and characteristics of the design Thermal Analysis and Test ESD Sensitivity Voltage Ramp Temperature Ramp The challenge for the high reliability user is to obtain the available data & determine its suitability to the application at hand Emission Microscopy Thermal Characterization 16

17 Product Acceptance For most low volume users, test and characterization of devices in their final packaged form is the only available option to empirically assess the suitability of the product to the intended application Screening and qualification tests must be designed with the objective of detecting failure mechanisms affecting infant mortality or life under normal operating conditions Plastic encapsulated microcircuits require special consideration in order to maintain the integrity of the components Construction analysis and destructive physical analysis are typically utilized to help the user in understanding the construction and workmanship of the product The challenge resides in the applied test conditions and interpretation of the resultant data. Utilization of standard MIL-STD type tests without consideration to material and design limitations may be detrimental Device characterization Cross sectional characterization 17

18 Simplified Parts Selection and Evaluation Flow Process Qualification Part Selection by designer Engineering Evaluation and Assessment Environmental Assessment Prior Test and Performance Data Manufacturing Process Evaluation Failure Modes and Mechanisms Procurement Specification Add to Approved Parts List Determination of Approval for Use Qualification Test Screening and Test Evaluation of Failure Product Acceptance Product Qualification 18

19 Special Tests and characterization may be necessary to determine suitability of the selected components to the intended space application Three main environmental conditions must be considered: Thermal Environment Operation at thermal condition beyond those experienced in normal operation on earth Mechanical shock environment Ability to withstand expected launch or acceleration environment Utilize analytical and experimental techniques Radiation Environment Characterization of radiation tolerance under the intended application Vibration Test & Characterization Thermal Characterization 19

20 Reliability Considerations

21 Definitions: Reliability: The Probability that an item Will Perform a Required Function Under Stated Conditions for a Stated Period of Time Probability - Described by Statistical Distribution Required Function - Defines Failure Time - Standard Unit of Measure (Seconds, Revolutions, etc.) Quality: A measure of the variance of a product from the desired state deviations in oxide thickness from the target value 21

22 Failure Classification Failures Can be Classified into two Groups: Catastrophic Failures The End of Component Life Complete Destruction of the Component Degradation Failures An Important Parameter Drifts so For From It s Original Value That The Component No Longer Functions Properly 22

23 Quantifying Reliability Traditional Reliability Determination Practices Submit a large Number of Samples to Actual Use Conditions Use Accelerated Life-Tests at Elevated Temperatures or stress Arrhenius equation R = A exp (-E a /KT) R = Rate of the Process A = Proportional Multiplier E a = Activation Energy, a Constant K = Boltzman s Constant (8.6x10-5 ev/k) Accelerated life test is a common practice for space qualification 23

24 Common Reliability Effects Die Oxide Breakdown Stress Induced Leakage Currents Hot Carrier Effects Electromigration Stress Voiding Mobile Ions Single Event Upset Electro Static Discharge Corrosion Junction Spiking... Package Adhesion Corrosion Moisture Cracking Fatigue Bond pad degradation... Process and product qualifications must address all known and applicable reliability effects 24

25 Reliability Considerations Drive Technology Changes Hot Carrier Injection Lightly Doped Drains (LDD) Electromigration Cu in Al Interconnect Junction Spiking Si in Al Interconnect 25

26 The Risk of Reliability Failures Has Increased As a Function of Device Scaling M transistors More potential failures sites 1,400M transistors Smaller defects will 0.25 micron features be fatal.050 micron features ± 250 mvolts allowable More sensitive to parameter drifts ± 60 mvolts allowable 26

27 Reliability Sensitivity to Processing Conditions Oxide Reliability starting material, oxide thickness, poly doping, plasma etching Hot Carrier Injection epi thickness, poly doping, channel length, LDD, S/D anneal Electromigration and Stress Voiding metalization, interlevel dielectrics, passivation Power Slump doping Profile, surface states, Source/Drain Ledge Reliability Mechanisms show a direct affect to changes in processing conditions 27

28 Reliability / Performance Trade End-of-life margins are often traded off against increased performance e.g., thinner gate dielectrics increase the speed of transistors which also increasing the risk of electrical breakdown The high volume, high turnover markets the industry focuses on have short lifetimes typically <5 years End-of-life wear-out is generally not seen in Integrated Circuits large end-of-life reliability safety margins 28

29 Yield and Reliability Early Failures (Defects) Constant Failure Rate End-of-Life (Intrinsic Failures) Failure Rate Higher Yields Impact Early Life Higher Performance Impacts Safety Margin Time Space Application Challenges Prediction of end-of-life reliability and understanding the yield-reliability relationship 29

30 Radiation Effects

31 Space Radiation Environments Deep Space Galactic cosmic rays heavy ions with extreme energies Solar flares (primarily protons and heavy ions) Trapped Radiation Belts Electrons up to 7 MeV Protons up to 400 MeV Galactic Cosmic Rays Solar Protons & Heavier Ions Trapped Particles Protons, Electrons, Heavy Ions Two basic effects: Permanent damage from the aggregate effect of many protons and electrons Localized spurious charges from heavy ions and protons 31

32 Environments of Earth-Orbiting Spacecraft Low-Earth Orbit (705 km, 98 º) Located below the edge of the proton belt Primary concerns are protons from the SAA and heavy ions Total dose ~ 20 krad (5-year mission) Geosynchronous Orbit (35.8 km) Located in outer electron belt Primary concerns are electrons from the belt and heavy ions Total dose ~ 50 krad (5-year mission) Outer Electron Belt South Atlantic Anomaly N Proton Belt Total dose values are approximate, and depend on shielding 32

33 Charge Generation by Heavy Ions and Protons Each particle produces an ionization track n p-substrate A few protons cause nuclear reactions Short-range recoil produces ionization Most protons pass through the device with little effect - + n + - p-substrate a) Heavy Ions (ionization b) Protons (nuclear reaction by each particle) needed to produce recoil) Short-duration charge pulse (~ ns) can upset digital and linear circuits Catastrophic damage can also take place for some device technologies 33

34 Military Environments High-intensity gamma-ray pulse Produces short-duration photocurrent pulse throughout a device structure Large circuit currents will collapse internal voltages, and may cause destruction Neutrons up to 14-MeV Produce permanent damage in semiconductors Reduces minority carrier lifetime (affect bipolar transistor gain) Also can affect carrier concentration 34

35 Approximate Total Dose Hardness Levels Silicon Compound semiconductors HFETs HBTs MESFETs Discrete transistors Memories Digital logic Linear amplifiers Light-emitting diodes Laser diodes (Only occurs with proton damage) (Only occurs with proton damage) Total Dose [krad] 35

36 Total Dose Effects in Silicon Devices Primary Effects: MOSFETs (including CMOS) Negative threshold shift in gate voltage Worse when positive voltage is applied to gate Depends on square of oxide thickness (less important for parts with small feature size) Increased leakage in field oxide Bipolar Devices (particular linear circuits) Gain is reduced Radiation damage can be far less under the low dose rate conditions in space compared to high dose-rate test conditions Silicon devices often rely on the high-quality properties of silicon dioxide. Trapped charge from radiation damage at surfaces can have a large effect. Compound semiconductors do not depend on an oxide layer and can isolate surface traps, reducing the importance of charge traps from radiation. 36

37 Total Dose Damage in a Linear Circuit This figure shows that very large changes can occur in a linear integrated circuit when it is tested at low dose rate compared to results at high dose rate. This is referred to as enhanced damage at low dose rate (ELDRs) This is a major concern for space systems. It only applies to bipolar linear circuits. The effect is due to the presence of thick oxides in the manufacturing process. Change in Input Offset Voltage (mv) 40 Analog Dev. OP rad/s rad/s rad/s (four-month test) rad/s Total Dose [krad(si)] 37

38 Effect of Proton Damage on a High-Frequency SiGe HBT Some degradation occurs at low collector currents There is essentially no degradation at the normal operating current level, even at such high total dose levels Pre-radiation Common-Emitter Current Gain x10 13 p/ cm 2 (9 Mrad) SiGe HBT 46 MeV Protons RF Bias Region I C (ma) 38

39 Effect of Total Dose Damage on an InP Heterojunction Bipolar Transistor III-V HBTs have very thin base regions, and are usually highly resistant to ionization damage. 100 Common-Emitter Current Gain Unexposed 56 Mrad 360 Mrad 620 Mrad V CE = 1.5V I C (ma) 10 39

40 Radiation Testing for Total Dose Cobalt-60 gamma rays are usually used Convenient, low cost irradiator Does not simulate displacement damage effects, only ionization Total dose damage is affected by bias conditions applied during irradiation The usual approach is to irradiate devices at several levels, making measurements between successive irradiations A room type irradiator. The cobalt-60 source is contained in a large lead enclosure, and raised during the irradiation period. Samples are placed near the throat of the irradiator 40

41 Radiation Testing for Displacement Damage This diagram shows an array of lightemitting diodes that are placed at the exit port of a proton accelerator The tests are expensive, because several people are required to operate the facility. Measurements are automated to reduce testing time. LED array Proton accelerator To bias network a) Irradiation Transition block LED array Detector array To measurement equipment To measurement equipment b) Measurements 41

42 Effect of Proton Damage on Optocouplers Optocouplers can be extremely sensitive to radiation damage Damage in the light-emitting diode is the cause I C I F CTR = I C IF In this example, the current transfer ratio degrades by a factor of 10 at 5 krad Very little damage occurs if the device is tested with gamma rays (displacement damage from protons is the mechanism) Change in Current Transfer Ratio (normalized) I F = 1 ma Optek Micropac Proton Fluence [p/cm2] 1x1011 2x1011 3x1011 4x1011 5x1011 Cobalt-60 gamma rays Equivalent Total Dose [krad (Si)] for Protons & Gamma Rays

43 Qualification Issues: Device Variability Many parts used in space are commercial devices, not specifically designed for radiation or for long-duration missions Data bases provide a guide for overall susceptibility, but radiation tests of specific lots may be required to ensure satisfactory performance The figure shows how the radiation damage in optocouplers from one manufacturer vary over a six-year time period. There is about a factor of 3 range in the proton fluence for 5X drop in current transfer ratio CTR (normalized) Lot 9518 Lot 9810 Lot 9838 Lot Proton Fluence (p/cm 2 )

44 Single-Event Upset Testing Devices are placed at the exit port of a cyclotron The cyclotron produces charged heavy ions The range of the ions is limited (they have much lower energies than ions in space) Usually the part lid must be removed to allow the ions to reach the active chip Some facilities require that the tests are done in a vacuum chamber The figure shows a complex test board that is being worked on in the open chamber of a cyclotron test facility Local instrumentation must be used to measure the response of devices to the ions. 44

45 Single-Event Upset Testing Pulses from the output of an optocoupler during tests at a heavy-ion facility. Special line drivers were used to allow the signals to be transmitted to an oscilloscope, located 30 feet away from the test devices. The purpose of most tests is to measure the cross section for particles with various values of linearenergy transfer N134 Optocoupler 6 5 Cross Section (cm 2 ) µm range 39 µm range Diode area Output Voltage (V) HP 6N134 Optocoupler 14 waveforms -0.5V trigger level LET (MeV-cm 2 /mg) Time (ns) 45

46 SEU Mitigation Several mitigation approaches can be used for SEU, including errordetection-and-correction (EDAC) Older 4-Mb DRAMs were used on the Cassini spacecraft Galactic cosmic rays produced about 307 errors per day in a 2.4 Gbit arra) The EDAC method could correct for all single-bit errors, and detect double-bit errors EDAC is more difficult for modern DRAMs because of complex upsets that scramble the chip, and the possibility of hard errors that remain after an ion strike. Total Error Cross Section (cm 2 ) Cross section for correctable and hard (uncorrectable errors) in a 64-Mb SDRAM Soft errors 0 20 Hard errors Hyundai 64-Mb SDRAM LET (MeV cm 2 /mg) 46

47 Single-Event Effects Comparison for Silicon and Compound Semiconductors Silicon Technology Very large scale circuits are used Circuits may upset in simple, correctable ways Complex upsets can also occur that crash parts, requiring reset Large numbers of circuits are used, increasing overall impact of singleevent upset effects CMOS technology is susceptible to latchup (often catastrophic) This is a major concern for space use Process and manufacturing changes can affect latchup susceptibility Compound Semiconductors Large scale circuits are rarely used Low hole mobility limits use of complementary logic Several specialized areas of importance: High-speed, high power devices (MESFETs and HFETs) Optoelectronics, particularly LEDs and laser diodes 47

48 Effects of Latchup on a CMOS Circuit Latchup is possible because parasitic bipolar transistors are present in most CMOS structures. They form a p-n-pn SCR that can be triggered on by heavy ions or protons, producing a large, localized current. Break in metallization caused by very high current from latchup Melting of silicon in highly localized region of current flow during latchup 48

49 Catastrophic Damage in High-Voltage SiC Diodes after Exposure to Heavy Ions This figure shows how the breakdown voltage in a 600-V SiC diode is affected when it is exposed to a low fluence of heavy ions. Extensive derating is necessary if this part is used in a radiation environment. 600 The damage mechanism may be related to the high defect density of silicon carbide Average breakdown voltage [V] CSD V, 1A rating LET (SiC) [MeV-cm 2 /mg] 49

50 Approximate Testing Costs Facility costs Total dose (~ $100 per hour) Proton accelerator (~ $700 per hour + travel time for experimenters) Overall test cost Total dose (~ $15 to $20k) Proton test (~$20 to $30k) Costs may be higher for parts that require elaborate measurements, or extensive analysis of results 50

51 Testing for Military Environments Neutrons Similar in concept to proton tests Protons and neutrons produce similar results for most devices, and are sometimes considered interchangeable Gamma Rays Tests are usually done at a flash X-ray The facility produces very intense, short-duration pulses The part(s) to be tested must be biased, usually with special instrumentation to measure currents and voltages A great deal of noise is introduced by the X-ray pulse Gamma ray test costs are higher, ~ $40 to $60k 51

52 Overview of Microelectronics Failure Mechanisms 52

53 The Evolution 1947: Ge transistor J. Bardeen, W. Brattain, W. Shockley 2007: 90nm Pentium M Processor on 300 mm silicon wafer 53

54 Common Failure Mechanisms in Silicon Ultra- Shallow Junctions Resistive contacts Junction spiking Sub-threshold leakage Punch through Gate Electrode Band bending B penetration (p channel) Drain induced barrier lowering V th control Ultra-thin Gate Insulator Direct tunneling gate currents Surface defects Stress induced leakage currents 54

55 Common Failure Mechanisms in GaAs Surface States Ohmic Contact Degradation Channel Degradation Gate Metal Sinking 55

56 Failure Mechanism Classifications Material-Interaction Induced Mechanisms Stress Induced Mechanisms Mechanically Induced Mechanisms Environmentally Induced Mechanisms 56

57 Material Interaction Induced Mechanisms Gate Metal Sinking Gate Au Contact Degradation Source Pt Drain Ti Channel Degradation GaAs Semi-insulating GaAs Surface State Effects 57

58 Gate Metal Sinking FIB Cross-sections of Control and Two Degraded Gate Locations Gate Sinking Caused by 4380 hours at 260ºC 58

59 Base Defects 0.2m E B B AlGaAs C crystalline defects in base TEM/FIB cross section of 2m emitter AlGaAs/GaAs HBT after beta degradation of ~30% 59

60 Surface States Effects Schematic cross section of a MESFET with Different surface charges. The gate-drain bias is the same for the two cases: (a) with low density of surface states Ds and (b) with high density of Ds. 60

61 Stress Induced Failure Mechanisms Electromigration Electrical Stress Hot Electron Trapping 61

62 Electromigration The movement of metal atoms along a metallic strip due to momentum exchange with electrons Depends on the temperature and the number of electrons Generally seen in narrow gates and in power devices where the current density is greater than 2x10 5 A/cm 2 Effect is observed both perpendicular to and along the source and drain contact edges and also at the interconnect of multilevel metallization. Depletion and accumulation of material in AuGeIn source and drain ohmic contacts induced by electromigration in a low-noise MESFET after life test. 62

63 Electrical Stress Airbridge Metal Deformation Caused By High Current >4 million A/cm 2, 175ºC, 1000 hours 63

64 ESD Damage On-Chip Damage as a result of Electro-Static Discharge 64

65 Electrical Overstress Electrostatic induced damage 65

66 Electrical Stress HBM ESD Damage in Capacitor (~300Volts) 50um x 50 um MIM Capacitor, 2000 Å Nitride 66

67 Hot Electron Trapping Under RF Overdrive, hot electrons are generated near the drain end of the channel where the electrical field is the highest. Electrons can accumulate sufficient energy to tunnel into Si3N4 passivation to form permanent traps. The traps can result in lower open-channel drain current and transconductance, and higher knee voltage, leakage current, and breakdown voltage. 67

68 Mechanically Induced Mechanisms Die Fracture Die Attach Voids Surface Defects Metal Voids 68

69 Die Fracture Die Crack Discovered After IR Reflow Simulation 69

70 Die Attach Voids Infra-red Image Showing Poor Die Attach 70

71 Surface Defects Metal B Possible Reliability Failure C Surface Defects have become a Very Critical Discriminator of Yield Definite Cause of Failure D E Detrimental Particle Size has Shrunk in response to Reduction of Feature Size A No Test or Reliability Failure 71

72 Air Bridge Voids 72

73 Via Metal Voids Incomplete Metal in Via 73

74 Fabrication Defects voids (arrows) reduced Al thickness (70%) 74

75 Lead Delamination Infra-red Image Showing lead Delamination and Poor Contact 75

76 Environmentally Induced Mechanisms Humidity Effects Hydrogen Effects Ionic Contamination 76

77 Humidity Effects 77

78 Hydrogen Effects in GaAs & InP Changes in peak transconductance,g m, and drain current at zero bias, I dss, of (a) InP HEMT and (b) GaAs PHEMT under nitrogen and 4% hydrogen treatment at 270 C 78

79 Hydrogen Effects - Example S ta nd a rd MES FET fa ilure Me c ha nis m in Nitro g e n Ac tiva tio n Ene rg y 1.8e V FET d e g ra d a tio n in 10% Hyd ro g e n Ac tiva tio n Ene rg y.93 to 1.0 e V 2 0 % De g ra da tio n % De g ra da tio n Me dian Time To Failure (Ho urs ) 79

80 Summary The environmental, reliability and operational challenges present a challenge to the application of commercially available microelectronics. Qualification of microelectronics for space application requires an understanding and consideration of the space environmental effects and relies on process qualification, products qualification and test under the intended use conditions. Prediction of end-of-life of a product requires an understanding of relationship between product yield and reliability. Space radiation effects are a major driver of acceptability of a product for reliable space application. All aspects of reliability and relevant known failure modes and mechanisms should be addressed prior to the insertion of the component in the application. 80